A wireless controlled robotic insect with ultrafast untethered running speeds

Insects attain high relative speeds and agile turning to evade risks or capture prey; for example, mites and tiger beetles exceed 170 BL s−1. These capabilities inspire insect‑scale legged microrobots, yet most prior systems achieve high performance only when tethered to external power and control. Integrating onboard power, control units, and task payloads typically causes large speed drops, limiting real‑world use. Existing microrobots adopt either walking gaits (e.g., Harvard HAMR) with complex actuation and diminished untethered speed, or running gaits using high‑frequency body bouncing (often soft robots) that degrade significantly under payload. The research question is how to design an insect‑scale legged microrobot that maintains ultrafast untethered speed and agile maneuverability even with payloads. This work proposes BHMbot, which uses impacts of independently driven rigid front legs and a body tilt to generate oblique bouncing, aiming to preserve high frequency bouncing under load for ultrafast untethered running and controllable trajectories.

Prior work shows legged gaits at small scales fall into walking (alternating strides, complex actuation; HAMR 4.5 cm reaches 11.4 BL s−1 tethered but 3.8 BL s−1 untethered; miniaturized HAMR-Jr faces untethered challenges) and running gaits (bouncing at high frequency via body deformation; soft microrobots can reach high tethered speeds, e.g., SEMR at 70 BL s−1 tethered). However, adding onboard power/control reduces speed dramatically (e.g., <2.5 BL s−1) because payloads damp body deformation. The literature indicates a gap in maintaining high relative speed and agility when untethered and carrying realistic payloads. Electromagnetic, piezoelectric, dielectric elastomer, and electrostatic actuation have been explored, each with trade‑offs in voltage, power density, and integration complexity.

Design: BHMbot (2.0 cm length untethered; core body 1.5 cm in prototypes) uses two electromagnetic linear actuators, each driving a front leg through a planar four‑bar transmission with a flexible hinge. Two passive rear legs and support frames complete the structure. Front legs are longer than rear legs to impose an upward body tilt angle for generating oblique reaction forces during leg ground impact. Actuators comprise a cantilever spring, NdFeB permanent magnet, and a copper hollow coil with added stainless‑steel baffles to reduce magnetic coupling. The actuator operates at low voltage (~1.2 V AC), obviating a high‑voltage booster. Actuation and gait: Alternating current drives magnet vibration; the four‑bar converts to leg swing. During the backward swing, the front leg pushes on the ground, producing an obliquely upward force that propels body bouncing with an aerial phase. Running speed v is modeled as f_bounce × L_bounce. Prototyping and tests: Multiple prototypes (10–25 mm body length) were fabricated (carbon fiber laminates, polyimide, adhesives; SCM for transmission; laser cutting for frames), and tested tethered across frequencies and currents (near resonance) with varying payload masses to establish speed trends, power, and optimization choices. Dynamic modeling and optimization: A planar two‑body model with a lumped payload was developed. The pin joint between body and front leg is modeled as a torsional spring‑damper driven by sinusoidal torque. Four generalized coordinates describe body vertical and pitch motions and leg angle. Parameters (coil–magnet gap z, cantilever width w_e, initial body tilt θ0, rear‑leg length l) were optimized to maximize running speed under set payloads. Optimal values without payload were near w_e≈1.4 mm, z≈1.5 mm; with w_e and z fixed, θ0≈9° and l≈7 mm maximized speed. Control electronics: A 600 mg wireless PCB (2.0×1.0 cm) integrates a Bluetooth MCU, dual Schmitt trigger, two H‑bridge drivers, two voltage regulators (1.2 V for drivers; 3.3 V for MCU/microphone), delivering two independent low‑voltage AC channels to the coils. Power from a 3.7 V, 50 mAh Li‑ion battery (800 mg). Frequency of each channel sets speed/turning; frequency difference controls turning radius and direction. Measurements: High‑speed imaging quantified bouncing phases, foot contact, COM motion, and duty cycles. Speeds and turning metrics measured on surfaces with measured friction coefficients (glass, wood, paper, plastic). Energy metrics captured via galvanometer for tethered tests and battery output for untethered tests. Slope and special surface (wet plastic, curved pipe) trials were performed. Application demonstrations included SOS sound detection (MEMS microphone), turbofan/turbojet engine passages, and quadrotor transport.

• Ultrafast untethered speed and agility: 2-cm BHMbot (1.76 g) achieves 17.5 BL s−1 linear speed and up to 65.4 BL s−2 relative centripetal acceleration (anticlockwise; 39.4 BL s−2 clockwise) on paper, with low‑radius rapid turns (e.g., ~0.7–1.0 cm radius for ~320–300° turns in 0.4 s with one channel near resonance and the other off).
• Energy metrics: Total Cost of Transport COT_T = 303.7 at measured battery output power Po = 1.77 W; measured actuator output power Pa = 5.62 mW yields actuation‑mechanism COT_M = 9.3. With a 50 mAh battery, maximum runtime ≈3 min.
• Payload-speed relationship: Unlike prior soft running microrobots, BHMbot’s speed increases with payload up to an optimal mass (m_op) then declines. For a 15 mm body length prototype, speed peaks at payload ~1200 mg (max 29.2 BL s−1 tethered); larger body lengths have larger optimal payloads (e.g., 20 mm: m_op≈2400 mg; 25 mm: m_op≈4800 mg). After integrating ~2 g payloads, optimized prototypes still achieved high speeds (e.g., 25 BL s−1 tethered with 2 g payload).
• Mechanism of performance under load: Payload reduces bouncing length but increases bouncing frequency; the product maintains or increases speed up to m_op. Duty cycle shifts show reduced aerial phase with payload, increasing effective power utilization.
• Optimized tethered performance: Prototype #3 (1.5 cm, 380 mg) reached 50 cm s−1 (33.3 BL s−1) tethered; Prototype #4 (same size) reached 25 BL s−1 tethered carrying a 2000 mg payload.
• Trajectory control: Wireless frequency control enabled precise circles (diameter 10.5 cm in 4.5 s), rectangles (11.5×14 cm in 8.0 s), letters “BUAA,” and 27 real‑time motion changes over a 73 cm irregular path through obstacles and tunnels (2.5 cm high, 4 cm wide).
• Surface and slope capability: Maximum relative speeds recorded across surfaces with μ≈0.161–0.372; robust turning accelerations on all. Climbed 6° slope at 6.5 BL s−1. Operated at ~10 BL s−1 on wet plastic with puddles and ~2.5 BL s−1 inside a 10 cm‑ID pipe.
• Comparative performance: Relative speed (17.5 BL s−1) surpasses many insects (e.g., common cockroach 13 BL s−1) and all reported untethered insect‑scale microrobots at similar or lower masses; turning agility (65.4 BL s−2) exceeds reported untethered robots up to 1 m length and insects like honey bees and cockroaches at similar scale.

The study addresses the core challenge of maintaining high locomotion performance at insect scale after integrating onboard power, control, and sensors. BHMbot’s design uses independent rigid front leg impacts and an intentional body tilt to create oblique bouncing, enabling a complementary relationship between bouncing length and frequency under payload. As payload increases from zero to m_op, the aerial phase shortens and bouncing frequency rises to match actuation frequency, improving effective power use and thus speed; beyond m_op, increased dissipation and reduced bounce length lower speed. Dynamic modeling guided optimization of key parameters (coil–magnet gap, cantilever width, body tilt, rear-leg length) to achieve high resonant frequencies and favorable contact dynamics, which translated to record untethered relative speed and centripetal acceleration at 2 cm body length. Comparisons with biological benchmarks and prior untethered robots show that the approach narrows the performance gap with insects and surpasses prior microrobots in both speed and turning agility. Energy analysis reveals system‑level efficiency constraints (high COT_T) dominated by electronics and actuator losses, while the actuation mechanism itself is comparatively efficient (low COT_M), indicating potential for further integration improvements.

This work introduces BHMbot, a 2‑cm wireless‑controlled legged microrobot that achieves ultrafast untethered running (17.5 BL s−1) and exceptional turning agility (up to 65.4 BL s−2). Key contributions include: a rigid‑leg running gait leveraging oblique impacts, dynamic modeling‑based structural optimization for high frequency bouncing under payload, and a minimal two‑channel low‑voltage control architecture enabling rich trajectory control. Demonstrations span complex trajectories, obstacle negotiation, wet and curved surfaces, slope climbing, SOS sound detection with a MEMS microphone, inspection within aero engines, and collaboration with a quadrotor for transport. Future work may focus on extending operational endurance (higher‑efficiency power electronics, improved actuators), enabling vertical mobility (e.g., climbing), integrating onboard vision (micro cameras) and autonomy, and further miniaturization and robustness for field deployment.

• Limited battery life (~3 minutes with 50 mAh) constrains mission duration.
• Inability to climb limits vertical mobility and access to elevated or complex 3D terrains.
• System‑level energy efficiency (COT_T ≈ 303.7) is reduced by high currents and actuator/electronics losses, despite efficient actuation mechanism (COT_M ≈ 9.3).
• Manufacturing and assembly tolerances introduce asymmetries, requiring small frequency offsets (30–40 Hz) for straight runs and affecting repeatability.
• Performance degrades beyond an optimal payload mass; speed–payload trade‑off depends on structural parameters and surface friction.
• Control relies on external computation and wireless commands; full onboard autonomy is not yet realized.